Baijiu blending method and system

Abstract

The invention discloses a Baijiu blending method and system. The method comprises the following steps: obtaining compound components and corresponding grades of base liquor of Maotai-flavor Baijiu; setting economic constraint requirements of combinatorial optimization; establishing a base wine component significance analysis model, and calculating a significance factor of each base wine component; screening corresponding base liquor components as basic factors of the combination by using the significance factors of the base liquor components, and establishing a base liquor combination optimization model based on the basic factors to obtain a candidate base liquor combination list corresponding to vinosity improvement; and reordering the candidate base liquor combination list obtained by combination optimization based on economic constraint requirements to obtain a base liquor combination list meeting the expectation, outputting the list for manual evaluation, and finally obtaining a base liquor blending scheme of the high-quality white spirit. According to the invention, the quality of the blending wine and the typing rating efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of Baijiu brewing, and particularly to a Baijiu blending method and system. Background Art

[0002] In the production process of Baijiu, blending is a crucial step among numerous processes. Before blending, it is necessary to classify and grade each Baijiu sample to be blended, and currently, the discrimination of the quality of base liquor almost entirely relies on sensory evaluation. Due to the complexity of the Baijiu production process and flow, blending Baijiu requires extremely high experience, professional qualities, and sensory sensitivity of the staff. Some enterprises cannot accurately and efficiently obtain the optimal blending ratio and plan due to problems such as the lack of sensory evaluation skilled personnel or bartenders and limited storage space for liquor. This only ensures the production volume while neglecting the liquor quality, seriously affecting the stability of product quality and the high-quality product rate. Improving the efficiency of classification and grading and efficiently utilizing the Baijiu samples to be blended is crucial for enhancing the efficiency of enterprises.

[0003] Currently, the academic and industrial circles mainly focus on the improvement of Baijiu production processes and production equipment, while there is less research on the theoretical methods of automated Baijiu blending. In existing related methods, such as the patent application publication number CN117150790A, titled "A Blending Method, Device and System for Maotai-flavor Baijiu", a linear programming algorithm is used to calculate the base liquor blending ratio of the finished liquor to be blended. In the patent application publication number CN116401959A, titled "An Optimization Method for Baijiu Blending Process Based on Deep Reinforcement Learning", a reinforcement learning algorithm is adopted to comprehensively consider the relationship between trace components and cost. However, the above existing technologies do not consider the importance degree among the physical and chemical indexes of Baijiu, resulting in poor model stability and generality. At the same time, the current existing technologies do not consider the sorting and screening of blending plans, which will affect the effect of the blending method and the input of human and material resources.

[0004] Therefore, it is necessary to provide a new Baijiu blending method and system to make up for the deficiencies of the existing technologies. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a Baijiu blending method and system, which can improve the quality of blended Baijiu and the efficiency of classification and grading.

[0006] The present invention solves the technical problems by adopting the following technical solutions:

[0007] A Baijiu blending method includes the following steps:

[0008] Step S1, obtaining the compound components and corresponding grades of Maotai-flavor Baijiu base liquor; setting the economic constraint requirements for combinatorial optimization;

[0009] Step S2, establish a significance analysis model for base liquor components, and calculate the significance factor for each base liquor component;

[0010] Step S3, use the significance factors of base liquor components to screen the corresponding base liquor components as the basic factors for combination, and based on this, establish an optimization model for base liquor combination to obtain a list of candidate base liquor combinations corresponding to the improvement of liquor quality;

[0011] Step S4, reorder the list of candidate base liquor combinations obtained by combination optimization based on economic constraint requirements to obtain a list of base liquor combinations that meet expectations, and output this list for manual evaluation, and finally obtain the base liquor blending scheme for high-quality liquor.

[0012] Furthermore, the economic constraint requirements in Step S1 include the following steps:

[0013] Step S1.1, estimate the costs and benefits of different grades of liquor bodies and the inventory content of different grades of base liquors;

[0014] Step S1.2, set the upper and lower limits of the blending ratios of various types of liquor.

[0015] Furthermore, in Step S2, the method of establishing a significance analysis model for base liquor components and calculating the significance factor for each base liquor component includes:

[0016] Step S2.1, screen the data features, delete the features with missing values exceeding the threshold, then fill in the missing values to obtain the processed data, and normalize the processed data;

[0017] Step S2.2, divide the data set, and use methods such as statistical machine learning to train a significance analysis model for base liquor components and verify the training effect of the model;

[0018] Step S2.3, design a corresponding attribution algorithm based on the category characteristics of the above model to obtain the significance factors corresponding to each liquor component.

[0019] Furthermore, Step S2 also includes: estimating the causal effect of different base liquor components on the liquor body quality using the significance factors corresponding to each liquor component, and the method is as follows:

[0020] Step S2.4, discretize the base liquor component data;

[0021] Step S2.5, establish a structural causal model, and estimate the corresponding causal effect by controlling confounding variables.

[0022] Furthermore, the method for establishing the base liquor combination optimization model in Step S3 includes the following steps:

[0023] Step S3.1, enhance the processed data using the significance factors;

[0024] Step S3.2: Define the distance metric between data, and define the combination target and the list of samples to be combined.

[0025] Step S3.3: Establish the base liquor combination optimization model.

[0026] Step S3.4: Solving the base liquor combination optimization model corresponds to a planning problem in the field of operations research optimization. Design a corresponding solution algorithm according to the characteristics of the established planning model.

[0027] Furthermore, the base liquor combination optimization model is as follows:

[0028]

[0029]

[0030] where \(w\) is the combination coefficient vector of the samples, \(x\) is the feature vector of the sample to be blended, the subscript represents the sample label, \(x_{m}\) m and \(w_{m}\) m are the \(m\)-th feature and the corresponding combination coefficient, and \(y\) is the blending target.

[0031] Furthermore, the method for obtaining the list of base liquors that meet the expectations in Step S4 includes the following steps:

[0032] Step S4.1: Establish a corresponding economic cost model according to the economic constraint requirements.

[0033] Step S4.2: Calculate the economic cost corresponding to the list of combination schemes.

[0034] Step S4.3: Re - sort the list of candidate base liquor combinations obtained by combination optimization according to the economic cost.

[0035] A Chinese liquor blending system for implementing the Chinese liquor blending method described in any one of the above.

[0036] A Chinese liquor blending method and system provided by the present invention have the following beneficial effects:

[0037] (1) The present invention adopts a data - driven algorithm, has little need for specific domain knowledge of liquor, can process Chinese liquors of various flavors and various production stages, and has strong versatility.

[0038] (2) The process of the present invention is clear, the algorithm has theoretical guarantee and high calculation efficiency.

[0039] (3) The present invention is simply deployed. Only by inputting the content of the physical and chemical indexes of the liquor sample, the combination calculation can be completed to achieve end - to - end.

[0040] (4) The present invention is applicable to small sample data, and as the data increases, the accuracy of the present invention can also increase accordingly. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0043] Reference Figure 1 , the present invention provides a method for blending Chinese liquor. By using the measurement data of volatile flavor components of Chinese liquor and the corresponding ratings, a significance analysis model of base liquor components is established and trained, and the corresponding significance factors are obtained. On this basis, an optimization model for base liquor combination is established, and a preliminary blending plan for the blended Chinese liquor is calculated. An economic cost model is established to reorder the above preliminary blending plan to obtain the final blending plan. Specifically, it includes the following steps:

[0044] Step S1: Obtain the compound components and corresponding grades of the base liquor of Maotai-flavor Chinese liquor; set the economic constraint requirements for combination optimization;

[0045] The economic constraint requirements include the following steps:

[0046] Step S1.1: Estimate the costs and benefits of liquor bodies of different grades and the inventory contents of base liquors of different grades;

[0047] Step S1.2: Set the upper and lower limits of the blending ratios of various liquors.

[0048] Step S2: Establish a significance analysis model of base liquor components and calculate the significance factors of each base liquor component;

[0049] The method for establishing a significance analysis model of base liquor components and calculating the significance factors of each base liquor component includes:

[0050] Step S2.1: Screen the data features, delete the features with missing values exceeding the threshold, then fill the missing values to obtain the processed data, and normalize the processed data;

[0051] Step S2.2: Divide the data set, and train a significance analysis model of base liquor components by means of statistical machine learning and other methods, and verify the training effect of the model;

[0052] Step S2.3: Design a corresponding attribution algorithm based on the category characteristics of the above model to obtain the salience factors corresponding to each Baijiu component.

[0053] Estimate the causal effect of different base liquor components on the quality of the liquor body using the salience factors corresponding to each Baijiu component. The method is as follows:

[0054] Step S2.4: Discretize the base liquor component data;

[0055] Step S2.5: Establish a structural causal model and estimate the corresponding causal effect by controlling confounding variables.

[0056] Step S3: Use the salience factors of the base liquor components to screen the corresponding base liquor components as the basic factors of the combination. Based on this, establish an optimization model for the base liquor combination to obtain a list of candidate base liquor combinations for improving the corresponding liquor quality;

[0057] The method for establishing the optimization model for the base liquor combination includes the following steps:

[0058] Step S3.1: Enhance the processed data using the salience factors;

[0059] Step S3.2: Define the distance metric between the data, define the combination objective and the list of samples to be combined;

[0060] Step S3.3: Establish an optimization model for the base liquor combination;

[0061] The optimization model for the base liquor combination is as follows:

[0062]

[0063]

[0064] where \(w\) is the combination coefficient vector of the samples, \(x\) is the feature vector of the sample to be blended, and the subscript represents the sample label, such as \(x\) m and \(w\) m are the \(m\)-th feature and the corresponding combination coefficient, and \(y\) is the blending objective.

[0065] Step S3.4: Solving the optimization model for the base liquor combination corresponds to a planning problem in the field of operations research optimization. Design a corresponding solution algorithm according to the characteristics of the established planning model.

[0066] The distance metric can select metric methods such as Euclidean distance, cosine distance, Mahalanobis distance, etc. The combined target can select representative wine samples with better quality. The samples to be combined can be adjusted according to the actual scenario. The planning model can select methods such as linear programming, fuzzy programming, quadratic programming, etc., and the solution method can select the corresponding algorithm for solution according to the function characteristics of the planning function and the feasible region. The setting of the constraint conditions can be adjusted according to the actual requirements.

[0067] Step S4, reorder the candidate base wine combination list obtained by the combined optimization based on the economic constraint requirements to obtain a base wine combination list that meets the expectations, and output this list for manual evaluation, and finally obtain the base wine blending plan for high-quality white liquor.

[0068] The method for obtaining the base wine combination list that meets the expectations includes the following steps:

[0069] Step S4.1, establish a corresponding economic cost model according to the economic constraint requirements;

[0070] Step S4.2, calculate the economic cost corresponding to the combination plan list;

[0071] Step S4.3, reorder the candidate base wine combination list obtained by the combined optimization according to the economic cost.

[0072] The present invention also provides a white liquor blending system for implementing the white liquor blending method.

[0073] The present invention studies the measurement data of the volatile flavor components of white liquor to establish and train a white liquor component analysis model, and obtains the corresponding significance factors. On this basis, a white liquor blending model is established, and the quality, utilization rate of white liquor and the workload of manual evaluation are improved through the synergistic effect and complementary advantages among the wine bodies. Finally, economic cost constraints are set, and the combination plan is screened and reordered in combination with the production cost to obtain the final plan. Simulation experiments and on-site expert evaluations verify the effectiveness of this method. The present invention has high calculation efficiency, a wide range of applicable scenarios, and can be flexibly adjusted according to the specific situation of the manufacturer; at the same time, the present invention improves the generality and stability of the model by calculating the significance factors of the white liquor components; in addition, the algorithm of the present invention is data-driven, has low requirements for prior knowledge in the field, and the accuracy of the model will continuously improve as the data volume increases.

[0074] Embodiment

[0075] In this embodiment, it is first necessary to measure the physical and chemical indicators of the liquor, and at the same time, it is required to conduct a grade assessment on a part of the samples. Based on this data, a significance analysis model of the base liquor components is trained to obtain the significance factors of each component. Based on the significance factors, the original data is enhanced, a base liquor combination optimization model is established, and a list of combination plans is obtained by solving. An economic benefit model is established based on the set economic constraint requirements, and the above list of combination plans is re-sorted to obtain the final blending plan. It mainly includes the following steps:

[0076] (1) Estimate the costs and benefits of liquors of different grades and the inventory contents of base liquors of different grades;

[0077] (2) Set the upper and lower limits of the blending ratios of various liquors and set constraints;

[0078] (3) Preprocess the data, screen the data features, delete the features with missing values exceeding 30%, fill the missing values with the mean value, and then divide the data set into a training set and a test set according to a 5-fold cross-validation ratio of 4:1. Train a random forest model with the training set and verify the training effect of the model with the test set. The mathematical model of the random forest is expressed as follows:

[0079]

[0080] where x is the training data, F 1 , …, F P are all physical and chemical indicators, and F b，1 , …, F b，K are all physical and chemical indicators obtained by decision tree sampling. Here, the maximum tree height is set to K = 3 and the number of subtrees is B = 3.

[0081] (4) Calculate the importance degree of different physical and chemical indicators based on the classification model established in step (1), and calculate the feature importance degree [φ1(f), …, φ1(f)] using the model-explainable random perturbation method. Then normalize it, and the formula is Scale the data features x i =(1 + α·φ i (f))x i . Among them, x i is the i-th column feature of the data set, and α is the control coefficient.

[0082] (5) Define the distance metric between a sample point and a set of sample points as dist(a, b) = ||a - b|| 2 . Establish the following combination optimization model:

[0083]

[0084]

[0085] Among them, w is the combined coefficient vector of the samples, x is the feature vector of the sample to be blended, and the subscript represents the sample label, such as x m and w m are the m-th feature and the corresponding combined coefficient, and y is the blending target.

[0086] (6) To solve the base liquor combination optimization model, the augmented Lagrangian multiplier method can be used to obtain the analytical solution. By adjusting the data source of the sample to be blended, different blending combination schemes can be obtained. According to the distance metric, n blending combination schemes can be screened and sorted for different schemes.

[0087] (7) Based on the economic constraint requirements set in step (1), an economic cost model is established. According to the cost and pricing of each liquor sample in the combination scheme list, an evaluation function is designed to obtain a scoring system.

[0088] (8) Reordering and screening: Based on the scoring system in step (7), the combination scheme list obtained in step (6) is scored and sorted, and the top m (m < n) combination schemes are selected as the final schemes.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for blending Chinese liquor, characterized in that, It includes the following steps: Step S1: Obtain the compound components and corresponding grades of the base liquor of Jiangxiang-type Baijiu; set the economic constraint requirements for combinatorial optimization; Step S2: Establish a significance analysis model for the base liquor components and calculate the significance factors of each base liquor component; Step S3: Use the significance factors of the base liquor components to screen the corresponding base liquor components as the basic factors of the combination. Based on this, establish a base liquor combination optimization model to obtain a list of candidate base liquor combinations corresponding to the improvement of liquor quality; Step S4: Reorder the list of candidate base liquor combinations obtained by combinatorial optimization based on the economic constraint requirements to obtain a list of base liquor combinations that meet the expectations, and output this list for manual evaluation, and finally obtain the base liquor blending scheme for high-quality Baijiu.

2. The baijiu blending method according to claim 1, characterized in that The economic constraint requirements in Step S1 include the following steps: Step S1.1: Estimate the costs and revenues of liquor bodies of different grades and the inventory contents of base liquors of different grades; Step S1.2: Set the upper and lower limits of the blending ratios of various liquors.

3. A method for blending Chinese liquor according to claim 2, wherein In Step S2, the method for establishing a significance analysis model for the base liquor components and calculating the significance factors of each base liquor component includes: Step S2.1: Screen the data features, delete the features with missing values exceeding the threshold, then fill in the missing values to obtain the processed data, and normalize the processed data; Step S2.2: Divide the data set, and use methods such as statistical machine learning to train a significance analysis model for the base liquor components and verify the training effect of the model; Step S2.3: Design a corresponding attribution algorithm based on the category characteristics of the above model to obtain the significance factors corresponding to each Baijiu component.

4. A method for blending Chinese liquor according to claim 3, characterized in that, Step S2 also includes: Estimating the causal effects of different base liquor components on the liquor body quality using the significance factors corresponding to each Baijiu component. The method is as follows: Step S2.4: Discretize the base liquor component data; Step S2.5: Establish a structural causal model and estimate the corresponding causal effects by controlling confounding variables.

5. A method for blending Chinese liquor according to claim 4, characterized in that, The method for establishing the base liquor combination optimization model in Step S3 includes the following steps: Step S3.1: Use the significance factors to enhance the processed data; Step S3.2: Define the distance metric between the data, define the combination objective and the list of samples to be combined; Step S3.3: Establish a base liquor combination optimization model; Step S3.4: The solution of the base liquor combination optimization model corresponds to a planning problem in the field of operations research optimization. Design a corresponding solution algorithm according to the characteristics of the established planning model.

6. The method for blending Chinese liquor according to claim 5, wherein, The base liquor combination optimization model is as follows: Among them, w is the combination coefficient vector of the samples, x is the feature vector of the sample to be blended, the subscript represents the sample label, x m and w m are the m-th feature and the corresponding combination coefficient, and y is the blending target.

7. A method for blending Chinese liquor according to claim 1 or 6, characterized in that The method for obtaining the list of base liquor combinations that meet the expectations in Step S4 includes the following steps: Step S4.1: Establish a corresponding economic cost model according to the economic constraint requirements; Step S4.2: Calculate the economic costs corresponding to the list of combination schemes; Step S4.3: Reorder the list of candidate base liquor combinations obtained by combinatorial optimization according to the economic costs.

8. A Chinese liquor blending system, characterized in that, It is used to implement the Baijiu blending method described in any one of claims 1-7.

Images